Pete Florence

Google DeepMind to Generalist AI Co-founder

Profile

FieldDetails
Current PositionGeneralist AI Co-founder & CEO
PreviousGoogle DeepMind Senior Research Scientist
PhDMIT (Computer Science)
AdvisorRuss Tedrake

Key Contributions

  • Dense Object Nets: Learning dense visual descriptors for objects (CoRL 2018 Best Paper)
  • PaLM-E: Embodied Multimodal Language Model co-author
  • RT-2: Vision-Language-Action Models co-author
  • Generalist AI Founding: Established general-purpose robot intelligence startup (2024)

Research Timeline

MIT PhD (Graduated 2019)

Advised by Russ Tedrake

YearWorkImpact
2018Dense Object NetsCoRL 2018 Best Paper Award
2019PhD Thesis: “Dense Visual Learning for Robot Manipulation”MIT Graduation

Google DeepMind (2019-2024)

Senior Research Scientist

YearWorkImpact
2020Self-Supervised CorrespondenceIEEE RA-L Best Paper Award
2023PaLM-EEmbodied Multimodal Language Model
2023RT-2Vision-Language-Action Models
2023Socratic ModelsICLR 2023

Generalist AI (2024-present)

Co-founder & CEO

YearWorkImpact
2024Co-founded Generalist AIFounded with Andy Zeng, Andrew Barry
2025NVentures InvestmentNvidia backing

Major Publications

Dense Descriptors

  • Dense Object Nets (CoRL 2018) - Best Paper Award
  • Self-Supervised Correspondence in Visuomotor Policy Learning (IEEE RA-L 2020) - Best Paper Award

Large Language Models + Robotics

  • PaLM-E: An Embodied Multimodal Language Model (2023)
  • RT-2: Vision-Language-Action Models (2023)
  • Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language (ICLR 2023)
  • Code as Policies: Language Model Programs for Embodied Control

Other Notable Works

  • DeepSDF (CVPR 2019 Best Paper Finalist)
  • Scaling Up and Distilling Down: Language-Guided Robot Skill Acquisition

Key Ideas

Dense Object Nets (2018)

Core: Learning consistent descriptors for every pixel on objects

Features:
- Invariant to viewpoint changes
- Correspondence between object instances
- Self-supervised learning

Applications:
- Robot grasping
- Object rearrangement
- Manipulation tasks

Impact:

  • Won CoRL 2018 Best Paper Award
  • Core research in robot visual representation learning
  • Pioneered dense correspondence application to robotics

Awards

  • CoRL 2018 Best Paper Award - Dense Object Nets
  • IEEE RA-L 2020 Best Paper Award - Self-Supervised Correspondence
  • CVPR 2019 Best Paper Finalist - DeepSDF

Generalist AI

In 2024, Florence left Google DeepMind to co-found Generalist AI with Andy Zeng and Andrew Barry.

Co-founders

  • Pete Florence (CEO) - formerly Google DeepMind
  • Andy Zeng (Chief Scientist) - formerly Google DeepMind
  • Andrew Barry (CTO) - formerly Boston Dynamics

Mission

The company aims to make general-purpose robots a reality, taking a new approach at the intersection of data, models, and hardware.

See Also